Proton released Lumo 1.1 with claims of faster, more detailed and more current responses while retaining its privacy-first design. The company positioned the assistant as an alternative to services that log conversations or reuse personal prompts for model training.
Proton reported a 200 percent improvement in complex reasoning, 170 percent better contextual understanding and a 40 percent increase in correct code generation. These were vendor metrics whose practical meaning depended on the underlying tests and user workloads.
The service encrypted conversations so Proton said it could not read them, did not save chats and did not use personal discussions for training. Its mobile applications were open source, allowing independent inspection of client-side behaviour.
Unlimited access and the strongest performance were associated with Lumo Plus, making subscription revenue an alternative to data-funded service. The product argued that users could pay directly rather than exchange behavioural information for a nominally free assistant.
The development should be read in its original publication context. It described the evidence, forecasts and vendor claims available at that time; later outcomes should not be projected backwards into promises that had not yet been tested.
A useful assessment separates demonstrated functions from proposed scale. It compares the new approach with a realistic baseline, including manual work, waiting time, existing infrastructure and the cost of correcting errors.
Long-term value requires more than launch performance. Data, models, organisations and external rules change, so version records, periodic checks and a named owner are needed to detect drift and decide whether the original case still holds.
Distribution also matters. An average gain can conceal predictable weakness for a language, site, workload or user group. Breaking results down and inviting feedback from people closest to the process makes expansion decisions more defensible.
Before scaling, leaders should map the complete delivery chain. Data sources, interfaces, cloud services, suppliers and human review can each become the limiting factor, even when the most visible model performs well. That map also exposes concentration, security and recovery requirements that a product demonstration rarely shows.
Financial analysis should follow outcomes rather than activity. More generated code, processed alerts or trained participants is not automatically more value. Measures such as completed tasks, avoided incidents, verified skills, lower correction effort and improved service quality give decision-makers a stronger basis for continuing investment.
Reversibility is equally important in a fast-moving market. Portable records, documented interfaces and clear stop criteria let an organisation change provider, narrow a deployment or return a decision to people when performance, pricing or policy changes. That flexibility is practical resilience, not a lack of commitment.
Version 1.1 Focused on Three Capability Gains
Reasoning, context and coding were the headline improvements, alongside faster and more current answers. The large percentages suggested material progress but lacked universal significance without task-level detail.
The practical importance was that users should test the assistant on their own repeated workflows and correction burden. A credible evaluation would connect the claim to representative outcomes, record exceptions and human intervention, and identify who could act when conditions changed. It should compare different workloads or user groups and document the resources required to sustain the result, preventing one headline metric from substituting for operational evidence. Teams would also need a baseline, a review schedule and a defined threshold for pausing or changing the system. Those controls would make progress measurable while preserving room to respond when costs, risks or user needs diverged from the original assumptions.
Encrypted Conversations Supported the Privacy Claim
Proton said employees could not read user chats and that conversations were neither retained nor used for training. That design reduced a common source of assistant-data exposure.
The practical importance was that privacy depended on the full architecture, including account metadata, retrieval and optional connected services. A credible evaluation would connect the claim to representative outcomes, record exceptions and human intervention, and identify who could act when conditions changed. It should compare different workloads or user groups and document the resources required to sustain the result, preventing one headline metric from substituting for operational evidence. Teams would also need a baseline, a review schedule and a defined threshold for pausing or changing the system. Those controls would make progress measurable while preserving room to respond when costs, risks or user needs diverged from the original assumptions.
Open-Source Apps Added Verifiability
Publishing mobile-client code let researchers inspect how the application handled conversations and communicated with the service. It did not automatically reveal every server-side component.
The practical importance was that open code strengthened trust when releases, builds and security reports remained easy to compare. A credible evaluation would connect the claim to representative outcomes, record exceptions and human intervention, and identify who could act when conditions changed. It should compare different workloads or user groups and document the resources required to sustain the result, preventing one headline metric from substituting for operational evidence. Teams would also need a baseline, a review schedule and a defined threshold for pausing or changing the system. Those controls would make progress measurable while preserving room to respond when costs, risks or user needs diverged from the original assumptions.
Subscription Changed the Business Incentive
Lumo Plus asked users to fund the service directly in exchange for unlimited access and higher performance. The commercial model aligned revenue more closely with serving customers than monetising their conversations.
The practical importance was that the long-term test was whether privacy and product quality remained sustainable at competitive pricing. A credible evaluation would connect the claim to representative outcomes, record exceptions and human intervention, and identify who could act when conditions changed. It should compare different workloads or user groups and document the resources required to sustain the result, preventing one headline metric from substituting for operational evidence. Teams would also need a baseline, a review schedule and a defined threshold for pausing or changing the system. Those controls would make progress measurable while preserving room to respond when costs, risks or user needs diverged from the original assumptions.
Across the announcement, technical capability was only one part of the system. Skills, access, integration, governance, cost and affected people could each determine whether the promised value survived outside a demonstration.
The durable lesson is to preserve traceability and choice. Organisations should document assumptions, keep consequential outputs reviewable, monitor downstream effects and retain a realistic exit route. Those controls allow ambitious AI work to improve without making early momentum an irreversible commitment.
